CHEMICAL STRUCTURE CHANGES OF COAL, CHAR, AND TAR

DURING DEVOLATILIZATION

Thomas H. Fletcher? and Ronald J. Pugmire$

?Chemical Engineering Dept., Brigham Young University, Provo, UT 84602

*Dept. of Chemical and Fuels Engineering, The University of Utah, Salt Lake City, UT 841 12

Keywords: coal, pyrolysis, 13C NMR

Introduction

Enormous progress has been made in coal pyrolysis research during the last decade.
Models of coal devolatilization have progressed from simple rate expressions based on total mass
release's to empirical relationships based on the elemental composition of the parent coal3 to
models that attempt to describe the macromolecular network of the ~ o a l . ~Measurements
-~ of
particle temperature during devolatilization have eliminated much of the controversy regarding
overall rates of devolatili~ation.~-'~In the last several years, advancements in chemical analysis
techniques have allowed quantitative investigations of the chemical structure of both coal and its
pyrolysis products, including the nature of the resulting char. A prominent research goal is to
accurately predict the rates, yields, and products of devolatilization from measurements of the
parent coal structure. This goal necessitates modeling the reaction processes on the molecular
scale, with activation energies that relate to chemical bond breaking rather than release of
products from the coal. 13C and ' H NMR spectroscopy have proven particularly useful in
obtaining average values of chemical structure features of coal, char, and tar.l1-I4 This paper
reviews experimental data regarding chemical structure features of coal, char, and tar during rapid
devolatilization, and how these data have impacted the development and input parameters for
devolatilization models. In particular, the relationship between pyridine extract yields and extract
yields predicted purely from NMR chemical structure data is discussed.

Parent Coal Structure

Coal consists of a macromolecular structure of fused aromatic rings connected by non-

aromatic bridges and loops. Although measurement of carbon aromaticity in coals has been
possible for a number of years, more detailed quantification of chemical features of coal structure
has become possible in the last five years. Advanced solid state 13C NMR techniques (CP/MAS
and dipolar dephasing), combined with carbon counting, have been used to determine the
chemical structure features of the Argonne Premium Coals.", l 5 In addition to carbon
aromaticity (fa,), the distinction between aromatic carbons with and without attachments (such as
hydrogen, carbon, or oxygen) is measured. A correlation was made to determine the average
number of aromatic carbons per fused aromatic cluster assuming circular rather than linear
catenation." The specification of the number of aromatic carbons per cluster (C,,) provides the
basis for the determination of many interesting chemical structure features. Probably one of the
most useful quantities is the number of attachments per aromatic cluster, referred to as the
coordination number (0+1), which is determined from the number of alkylated (f,q and phenolic
(faq attachments to aromatic carbons, as follows:

In addition to the total number of attachments per cluster, it is possible to determine the
fraction of total carbons in methyl and methoxy groups. This quantity identifies methyl groups,
which, t o a first approximation, are considered to be the only chain terminators. This allows
quantification of the total attachments per cluster into chains that terminate (i.e., attachments with
methyl groups, referred to as side chains) and chains that connect to other aromatic clusters
(referred to as bridges and loops). Since this is a carbon counting method, there may be some
discrepancy in that oxygen bridges are not counted. The fraction of attachments that are bridges
between clusters (p) is determined as follows:

and (1-p) is the fraction of attachments that are side chains. Orendt, et al.I5 showed that the
coordination number varies from 3.9 to 5.6 for the Argonne premium coals, while the fraction of
attachments existing as bridges and loops (po) ranged from 0.49 to 0.74. This means that on
average, each aromatic cluster is connected to other aromatic clusters at p(o+l) or 2.5 to 3.6
points. In contrast, a long chain-like polymer with no crosslinks contains 2.0 connecting bridges

108
 per aromatic cluster. In general, the more crosslinks per aromatic cluster (ie., as the coordination
number increases beyond 2.0). the harder it is for a material to thermally decompose.
Once the number of aromatic carbons per cluster and the number of total attachments per
cluster are determined, the average cluster molecular weight (Mcl, including side chains and one-
half the bridges and loops) can be determined using the elemental carbon composition (xc) and
the molecular weight of carbon (Mc):~,I '

This corresponds to the size of the average monomer unit in the coal macromolecule. The
average molecular weight of a side chain (Mg) can also be determined:

(4)

Values of cluster molecular weights (M,I) in parent coals determined using this method
range from 270 to 410 amu, with no clear trend with coal rank. Side chain molecular weight (Mg)
show a clear trend with coal rank, ranging from 12 amu for the high rank coals to 52 for the low
rank coals (see Fig. 1). These average values of Mg roughly correspond to methyl groups (15
amu) and carboxylic acid groups (45 amu), respectively. It must be remembered, however, that
these values represent weighted average of many types of attachments.

Structure of Pyrolysis Products (Chars and Tars)

Char and tar samples were obtained as a function of residence time in a devolatilization
experiment (1250 K, 2 x 104 Ws in nitrogen) and 'ust subsequent to devolatilization in a laminar
flame-fired experiment (1500 K, 5 x 104 Ws).I4 Quantitative measurements of chemical
structure were performed on the coals and chars using the 13C NMR techniques described above.
Results show that the chemical structures of fully-devolatilized chars are very similar, even
though a wide diversity is seen in the parent coal structures. For example, the average cluster
molecular weights of the fully-pyrolyzed chars span a range of only 50 amu, in contrast to the
span of 150 amu observed in the parent coals. Side chain molecular weights of the fully-
pyrolyzed chars span a narrow range from 11 to 18 amu, which contrasts even more with the
parent coal data (see Fig. 1). The change in side chain molecular weight is most dramatic for the
low rank coals, corresponding to the release of large amounts of aliphatic material as light gases.
The similarity in chemical structure of fully-devolatilized coal chars suggests that differences in
measured heterogeneous char reactivities may be influenced primarily by physical structure. l4

Use of Chemical Structure Features in Devolatilization Models

The use of statistics applied to polymer chains was applied to coal devolatilization by Niksa
and Kerstein.16 Devolatilization models have evolved to use network structures that represent
coal as aromatic clusters connected by labile bridge material!-6 A summary of these three
models was recently published." A non-linear relationship exists between the number of intact
labile bridges and the amount of clusters disconnected from the "infinite" lattice structure.18
Closed-form solutions have been formulated to relate the breakup of the lattice structure to the
bridge population and the initial characteristics of the lattice. Both straight chain lattices5 and
Bethe lattices6. have been used to represent the coal macromolecular structure.
The three coal devolatilization models utilize the chemical structure features available from
solid-state NMR analyses in different ways. All of the models use or reference the number of
aromatic carbons per cluster (ie., MWcluster). The number of attachments per cluster (a+l)are
used directly in the CPD and FG-DVC models, but straight chains are used in FLASHCHAIN.
The molecular weight per side chain (MWs) is used directly in the CPD model, and directly
impacts the light gas yield. In the FG-DVC model, the light gas species evolution is specified
based on empirical fits of light gas yields in TG-FTIR experiments, and therefore does not link
the light gas evolution directly to the lattice structure of the parent coal. In FLASHCHAIN, the
molecular weights of the bridge material are much larger than the NMR data, and a correction
factor is used to specify the final molecular weight of the light gas.
The direct use of NMR data to specify the coal-dependent parameters is illustrated in Fig. 2,
which compares predicted tar and total volatiles yields for sixteen coals during devolatilization at
heating rates from 1 to 10,OOO Us agree with measured values.14 For most coals, no adjustable
parameters were used to tune the predictions to match the yields. However, one adjustable
parameter was used for lignite to represent early crosslinking, while this same parameter was
adjusted to represent stable bi-aryl bridges in high rank coals (Le., Iv bituminous). In these
predictions, the total gas yield was calculated without the use of yield factors from previous
pyrolysis experiments.
The initial fraction of the labile bridges that are intact specifies the connectivity of the
lattice. The initial lattice is generally not fully connected, and a certain amount of free or

109
i
I
disconnected material is predicted by the lattice statistics. This material is generally thought to
correspond to the solvent extracts from unreacted coals. However, there seems to be a
discrepancy between the amount of "extracts" predicted by using the solid state NMR data and
measured pyridine extract yields. Figure 3 shows pyridine extract yields from the Argonne
Premium coals measured by Fletcher, et al.I9 Also shown are the predicted amounts of
unattached material (Le., extract) based on a Bethe lattice with coordination number and fraction
of intact bridges taken directly from NMR measurements using the CPD model. The measured
pyridine extract yields are as high as 26% for the Pittsburgh #8 coal, while the predicted yields
for this coal are significantly lower.
Possib!e xssons for the lack of agreement between measured pyridine extracts and
predictions based on NMR structural data are:

(i) errors in the NMR measurements and their interpretation

(ii) the use of average NMR structural data rather than distributions
(iii) the use of Bethe lattices as approximations of coal molecular structure
(iv) the representation of the extract and the residue by the same latticc stiucture, even
though they are chemically different

Of these hypotheses, (iv) seems to be the most rationale explanation for the disagreement
shown in Fig. 3. For example, the total number of attachments per cluster (a+l)determined by
13C NMR in the pyridine extracts is an average of 15% lower than in the corresponding residue,
as shown in Fig. 4 for the Argonne Premium coals.19 The number of aromatic carbons per
cluster in the pyridine extracts is also consistently lower than in the corresponding residues, and
the result is lower molecular weights per cluster in the extracts (Fig. 5). The number of bridges
and loops per cluster is also lower in the extracts than in the residues, although the molecular
weight per side chain in the extracts is similar to that in the corresponding residues. The
differences in average chemical structure features between the extracts and residues seems to
confirm hypothesis (iv) above; the extract is chemically different from the residue and should not
be treated with the same lattice structure. These results obviously have the most impact for those
coals with high pyridine extract yields.

Conclusion

Coal pyrolysis research has progressed to the point that measurements of the chemical
structure features of parent coals is useful in current devolatilization models. Chemical structure
features of particular importance to modeling efforts seem to be the number of aromatic carbons
per cluster, the number of attachments per cluster, and the total molecular weight per cluster
(including attachments). Most current models are able to describe tar and gas yields as a function
of time, temperature, heating rate, coal type, and pressure, although the network models of
devolatilization appear to have the closest ties with coal structure. However, the models
selectively use chemical structure information in order to attain agreement with measured tar and
gas yields, while other pertinent data are ignored. One such example is that the models use as
input parameters either measured pyridine extract yields o r the fraction of intact connecting
bridges. Chemical structure features of pyridine extracts do not match those of the corresponding
residue, suggesting that models should treat extracts as a different chemical than the coal. The
challenge for the models is to reduce the number of empirical parameters by using chemical
structure information.

Acknowledgments

This work was sponsored by the Advanced Combustion Engineering Research Center at
Brigham Young University and the University of Utah. Funds for this center are received from
the National Science Foundation, the State of Utah, 31 industrial participants, and the U. S.
Department of Energy. The authors also acknowledge the contributions of Dave Grant and Mark
Solum at the University of Utah who were key participants in much of the work reviewed here.

References

1. Anthony, D. B., J. B. Howard, H. C. Hottel and H. P. Meissner 15th Symposium

(International) on Combustion; The Combustion Institute, Pittsburgh, P A 1974;pp 1303-
1317.
2. Kobayashi, H.,J. B. Howard and A. F. Sarofim 16th Symposium (International) on
Combustion; The Combustion Institute, Pittsburgh, PA: 1976;pp 41 1-425.
3. KO,G. H., D. M. Sanchez, W. A. Peters and J. B. Howard 22nd Symposium (International)
on Combustion; The Combustion Institute, Pittsburgh, PA: 1988;pp 115-124.
4. Fletcher, T. H.,A. R. Kerstein, R. J. Pugmire and D. M. Grant Energy and Fuels 1992.6,
414.
5. Niksa, S.Energy and Fuels 1991.5.673-683,
110
 6. Solomon, P. R., D. G.Hamblen, R. M. Carangelo, M. A. Serio and G.V. Deshpande
Energy and Fuels 1988,2,405-422.
7. Solomon, P. R., M. R. Serio, R. M. Carangelo and J. R. Markham Fuel 1986.65, 182-193.
8. Fletcher, T. H. Combustion Science and Technology 1989,63,89.
9. Fletcher, T. H. Combustion and Flame 1989, 78,223.
10. Solomon, P. R., T. H. Fletcher and R. J. Pugmire Fuel 1993,72,587-597.
11. Solum, M. S., R. J. Pugmire and D. M. Grant Energy and Fuels 1989,3, 187.
12. Fletcher, T. H., M. S. Solum, D. M. Grant, S . Critchfield and R. J. Pugmire 23rd Symposium
(Infernational)on Combustion; The Combustion Institute, Pittsburgh, PA: 1990; pp 1231.
13. Pugmire, R. J., M. S. Solum, D. M. Grant, S . Critchfield and T. H. Fletcher Fuel 1991, 70,
414.
14. Fletcher, T. H., M. S. Solum, D. M. Grant and R. J. Pugmire Energy and Fuels 1992,6,
643-650.
15. Orendt, A. M., M. S. Solum, N. K. Sethi, R. J. Pugmire and D. M. Grant In Advances in
Coal Spectroscopy; H. L. C. Meuzelaar,Ed.; Plenum Press: New York, 1992 pp 215-254.
16. Niksa, S. and A. R. Kerstein Combustion and Flame 1986,66,95-109.
17. Smith, K. L., L. D. Smoot and T. H. Fletcher In Fundamentals of Coal Combustion for
Clean and EIficien?Use; L. D. Smoot, Ed.; Elsevier: New York, 1993; pp 131-298.
18. Grant, D. M., R. J. Pugmire, T. H. Fletcher and A. R. Kerstein Energy and Fuels 1989,3,
175-186.
19. Fletcher, T. H., S. Bai, R. J. Pugmire, M. S. Solum, S. Wood and D. M. Grant Energy and
F W ~ S1993, 7,734-742.

601 I I I I I I
-
40-

-
20 -

-
01 I I I I I
0 10 20 30

Oxygen Content of Parent Coal (% daf)

Figure 1. Average molecular weight of side chains attached to aromatic clusters in unreacted
\ coals and fully-devolatilized chars. Oxygen content of the parent coal is used as a
rank indicator. Lines represent linear correlations of the data (see Fletcher et aI.l4)
\

40 -
-
-
-
0 10 20 30 40 50 60 70

Measured Yield

Figure 2. Comparison of predicted tar and total volatiles yields from NMR structural data
versus measurements. Data are from sixteen coals during pyrolysis over a range of
heating rates (see Fletcher et aL4)

111
 PrediOed lrom NMR

-e
...
X

Figure 3. Measured pyridine extract yields compared with predictions of unconnected

fragments in a Bethe lattice using I3C NMR characterizations of average chemical
structure (see Fletcher, et a1.19).

Figure 4. Coordination numbers ((3+1) determined from I3C NMR analyses of pyridine
extractions of the Argonne premium coals (see Fletcher, et al.l9).

E -n g m8IdU
8 *-

Figure 5. Average molecular weight per cluster determined from 13C NMR analyses of
pyridine extractions of the Argonne premium coals (data from Fletcher, et al.l9).

112